Overflow out of margin at inline reference citation

Question

In my writing, I see that sometimes the text doesn't follow the margin imposed when a reference is added a the end of the word.

Here is an example:

The associated code is (I broke the line to make is more readable):

\documentclass[12pt, english, onehalfspacing, headsepline]{MastersDoctoralThesis}
\begin{document}
        Region-growing methods are particularly appropriate to segment large homogeneous connected regions
     and have been employed for numerous segmentation tasks in medical imaging using MRI such as ventricular
     blood pool segmentation\Cup{muhlenbruch2006global} or LA segmentation\Cup{zhu2013automatic}.
     However, region growing approaches are sensitive to noise which can impair correct pixel annotation
     and lead to holes in the region of interest.
     Moreover, region growing methods still require manual tuning such as selecting
     an appropriate seeding location and defining efficient similarity criterion
     to prevent the algorithm to "leak" outside of the region of interest.
\end{document}

With:

\newcommand\Cup[1]{\textsuperscript{\cite{#1}}}

Is there a way to improve the situation ?

EDIT: I am using a template from Overleaf:

% Masters/Doctoral Thesis 
% LaTeX Template
% Version 2.5 (27/8/17)
%
% This template was downloaded from:
% http://www.LaTeXTemplates.com
%
% Version 2.x major modifications by:
% Vel (vel@latextemplates.com)
%
% This template is based on a template by:
% Steve Gunn (http://users.ecs.soton.ac.uk/srg/softwaretools/document/templates/)
% Sunil Patel (http://www.sunilpatel.co.uk/thesis-template/)
%
% Template license:
% CC BY-NC-SA 3.0 (http://creativecommons.org/licenses/by-nc-sa/3.0/)

The other issues seem similar:

I imagine that the second example is due to the use of $\pm$. However, I am not too sure. Moreover, I don't know why the third one appears.

Here is the related code:

Misalignment 2:

(Dice score 93.67$\pm$0.2\% vs 92.91$\pm$0.38\%, MSD 0.689$\pm$0.107mm vs
 0.791$\pm$0.047mm, \textit{p}-value < 0.01). On the other hand, the application 
of solely histogram augmentations (gamma, CLAHE and add) (purple bars in Figure 
\ref{data_aug_final} A and B) significantly decrease the accuracy scores (Dice 
score 92.23$\pm$0.6\% vs 92.91$\pm$0.38\%, MSD 0.917$\pm$0.129mm vs 
0.791$\pm$0.047mm,\textit{p}-value < 0.01). Even if removing the add data 
augmentation improves the results, this

Misalignment 3:

    Using \textit{histogram augmentations} only the modification of the contrast
 using gamma showed improvements on the Dice score (93.06$\pm$0.51\% vs 
92.93$\pm$0.38\% using no data augmentation). No Dice score modification was 
yielded by using CLAHE (92.96\% vs 92.93\%) and "add" worsened the results lower 
the Dice score (92.66\% vs 92.93\%). For the MSD metrics, only the CLAHE improved 
the final score (0.758$\pm$0.06 vs 0.767$\pm$0.08mm) while "add"

Is there a package that I am not using correctly ?

I used these packages:

\usepackage[utf8]{inputenc} % Required for inputting international characters
\usepackage[T1]{fontenc} % Output font encoding for international characters
\usepackage{tocbibind}
\usepackage{mathpazo} % Use the Palatino font by default
\usepackage[backend=biber,style=ieee]{biblatex}
\usepackage[font=singlespacing, font=small]{caption}
\usepackage[autostyle=true]{csquotes}
\usepackage{graphicx}
\usepackage{wrapfig}
\usepackage{textcomp}
\usepackage{booktabs}
\usepackage{multirow}
\usepackage{siunitx}
\usepackage{amsfonts, amsmath, amsthm, amssymb}
\usepackage{color,soul}
\usepackage{float}
\usepackage{array}

EDIT 2: These later issues were fixed using the microtype package as \usepackage{microtype} in the "package definition section". However, some references are still not quite tidy.

Example:

Is there another/special package for it ?

Kind regards,

Answer 1

Your macro \Cup creates the previous word impossible to use hyphenation, because box is appended immediately to the word. This complicates the paragraph breaking, so overfull \hboxes are more probably.

Define

\def\Cup#1{\nobreak\textsuperscript{\cite{#1}}}

(or use \newcommand if you insist to the LaTeX notation:). The \nobreak macro inserts penalty between the word and the box constructed by \textsuperscript. The paragraph breaking algorithm accepts the previous word really as word in such case and hyphenation is allowed. So, segmentation can be divided by the rule seg-men-ta-tion if English hyphenation patterns are used.